解决安全模型更新时误报率不一致问题，确保预测分数稳定可靠。
Full-range Binary Classifier Calibration for Stable Model Updates in Production
- 基于现有校准技术，针对全范围误报率曲线进行统一校准
- 在0.1%至10%误报率区间内相对误差不超过2.3%
- 模型体积小（<200KB），适用于大规模生产部署
运行在对抗环境中的检测模型面临快速漂移的恶意分布，而良性分布相对稳定，因此需持续重训和重新部署以应对新威胁。但重训常导致输出预测分数变化，破坏下游系统的使用。针对此类安全模型，需要在所有输出值上保持一致的误报率（FPR），而传统概率校准方法仅关注类别概率而非FPR约束。本文提出一种基于现有校准原语的方法，可对整个FPR曲线进行校准，使不同部署下的分数具备一致的FPR含义。在独立测试集上，当FPR从10%降至0.1%时，观测到的相对FPR误差最大为2.3%，在0.01% FPR时为7.2%。校准后的模型文件大小始终低于200 KB，且在1K至10M个良性样本的不同校准集上均表现稳定。
原文摘要 · Abstract (English)
Detection models running in adversarial environments face a malicious distribution that drifts rapidly while the benign distribution stays comparatively stable, so teams retrain and redeploy constantly to stay ahead of new threats. Retraining tends to change the output prediction scores, which breaks downstream users of the model. For these security-oriented models we need consistent false-positive rate (FPR) across all output values, whereas standard probability-calibration methods target class probability rather than an FPR contract. We introduce a method built on top of existing calibration primitives that targets the whole FPR curve, giving scores a consistent FPR meaning across deployments. On one held-out split, the observed relative FPR error was at most 2.3% from 10% down to 0.1% FPR and 7.2% at 0.01% FPR. The shipped artifact remains under 200 KB in measurements across calibration sets from 1K to 10M benign samples.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。